[PATCH] page-types: decode flags directly from command line

[PATCH] page-types: decode flags directly from command line

[Alex Chiang]

Teach page-types to decode page flags directly from the command
line.

Why is this useful? For instance, if you're using memory hotplug
and see this in /var/log/messages:

	kernel: removing from LRU failed 3836dd0/1/1e00000000000400

It would be nice to decode those page flags without staring at
the source.

Example usage and output:

linux-2.6/Documentation/vm$ ./page-types -d 0x1e00000000000400
             flags	page-count       MB  symbolic-flags			long-symbolic-flags
0x1e00000000000400	         1        0  __________B_______________________buddy
             total	         1        0

Signed-off-by: Alex Chiang <achiang@hp.com>
---
 page-types.c |   26 ++++++++++++++++++++++++--
 1 file changed, 24 insertions(+), 2 deletions(-)
---
diff --git a/Documentation/vm/page-types.c b/Documentation/vm/page-types.c
index 3ec4f2a..a55c624 100644
--- a/Documentation/vm/page-types.c
+++ b/Documentation/vm/page-types.c
@@ -674,6 +674,7 @@ static void usage(void)
 	printf(
 "page-types [options]\n"
 "            -r|--raw                  Raw mode, for kernel developers\n"
+"            -d|--decode  flags        Decode a single page's flags\n"
 "            -a|--addr    addr-spec    Walk a range of pages\n"
 "            -b|--bits    bits-spec    Walk pages with specified bits\n"
 "            -p|--pid     pid          Walk process address space\n"
@@ -682,10 +683,12 @@ static void usage(void)
 #endif
 "            -l|--list                 Show page details in ranges\n"
 "            -L|--list-each            Show page details one by one\n"
-"            -N|--no-summary           Don't show summay info\n"
+"            -N|--no-summary           Don't show summary info\n"
 "            -X|--hwpoison             hwpoison pages\n"
 "            -x|--unpoison             unpoison pages\n"
 "            -h|--help                 Show this usage message\n"
+"flags:\n"
+"            0x0000000000000400        A single page's flags, e.g.\n"
 "addr-spec:\n"
 "            N                         one page at offset N (unit: pages)\n"
 "            N+M                       pages range from N to N+M-1\n"
@@ -884,12 +887,28 @@ static void parse_bits_mask(const char *optarg)
 	add_bits_filter(mask, bits);
 }
 
+static void decode_flags_and_exit(const char *optarg)
+{
+	uint64_t flags;
+
+	flags = parse_number(optarg);
+
+	opt_list = 0;
+	opt_hwpoison = 0;
+	opt_unpoison = 0;
+
+	add_page(0, 0, flags);
+	show_summary();
+
+	exit(0);
+}
 
 static struct option opts[] = {
 	{ "raw"       , 0, NULL, 'r' },
 	{ "pid"       , 1, NULL, 'p' },
 	{ "file"      , 1, NULL, 'f' },
 	{ "addr"      , 1, NULL, 'a' },
+	{ "decode"    , 1, NULL, 'd' },
 	{ "bits"      , 1, NULL, 'b' },
 	{ "list"      , 0, NULL, 'l' },
 	{ "list-each" , 0, NULL, 'L' },
@@ -907,7 +926,7 @@ int main(int argc, char *argv[])
 	page_size = getpagesize();
 
 	while ((c = getopt_long(argc, argv,
-				"rp:f:a:b:lLNXxh", opts, NULL)) != -1) {
+				"rp:f:a:b:d:lLNXxh", opts, NULL)) != -1) {
 		switch (c) {
 		case 'r':
 			opt_raw = 1;
@@ -924,6 +943,9 @@ int main(int argc, char *argv[])
 		case 'b':
 			parse_bits_mask(optarg);
 			break;
+		case 'd':
+			decode_flags_and_exit(optarg);
+			break;
 		case 'l':
 			opt_list = 1;
 			break;

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] page-types: decode flags directly from command line

[Wu Fengguang]

Hi Alex,

On Wed, Nov 04, 2009 at 06:54:41AM +0800, Alex Chiang wrote:
> Teach page-types to decode page flags directly from the command
> line.

This is good feature to have - I considered adding it, too ;)

> 
> Why is this useful? For instance, if you're using memory hotplug
> and see this in /var/log/messages:
> 
> 	kernel: removing from LRU failed 3836dd0/1/1e00000000000400
> 
> It would be nice to decode those page flags without staring at
> the source.

In fact it's more than decode - encoding is also possible with the
_same_ code! So maybe "-d" and help message will not be all that
appropriate.

> Example usage and output:
> 
> linux-2.6/Documentation/vm$ ./page-types -d 0x1e00000000000400
>              flags	page-count       MB  symbolic-flags			long-symbolic-flags
> 0x1e00000000000400	         1        0  __________B_______________________buddy
>              total	         1        0

The output is a bit redundant - so does the code. Could you simplify
them a bit?

Thanks,
Fengguang

> Signed-off-by: Alex Chiang <achiang@hp.com>
> ---
>  page-types.c |   26 ++++++++++++++++++++++++--
>  1 file changed, 24 insertions(+), 2 deletions(-)
> ---
> diff --git a/Documentation/vm/page-types.c b/Documentation/vm/page-types.c
> index 3ec4f2a..a55c624 100644
> --- a/Documentation/vm/page-types.c
> +++ b/Documentation/vm/page-types.c
> @@ -674,6 +674,7 @@ static void usage(void)
>  	printf(
>  "page-types [options]\n"
>  "            -r|--raw                  Raw mode, for kernel developers\n"
> +"            -d|--decode  flags        Decode a single page's flags\n"
>  "            -a|--addr    addr-spec    Walk a range of pages\n"
>  "            -b|--bits    bits-spec    Walk pages with specified bits\n"
>  "            -p|--pid     pid          Walk process address space\n"
> @@ -682,10 +683,12 @@ static void usage(void)
>  #endif
>  "            -l|--list                 Show page details in ranges\n"
>  "            -L|--list-each            Show page details one by one\n"
> -"            -N|--no-summary           Don't show summay info\n"
> +"            -N|--no-summary           Don't show summary info\n"
>  "            -X|--hwpoison             hwpoison pages\n"
>  "            -x|--unpoison             unpoison pages\n"
>  "            -h|--help                 Show this usage message\n"
> +"flags:\n"
> +"            0x0000000000000400        A single page's flags, e.g.\n"
>  "addr-spec:\n"
>  "            N                         one page at offset N (unit: pages)\n"
>  "            N+M                       pages range from N to N+M-1\n"
> @@ -884,12 +887,28 @@ static void parse_bits_mask(const char *optarg)
>  	add_bits_filter(mask, bits);
>  }
>  
> +static void decode_flags_and_exit(const char *optarg)
> +{
> +	uint64_t flags;
> +
> +	flags = parse_number(optarg);
> +
> +	opt_list = 0;
> +	opt_hwpoison = 0;
> +	opt_unpoison = 0;
> +
> +	add_page(0, 0, flags);
> +	show_summary();
> +
> +	exit(0);
> +}
>  
>  static struct option opts[] = {
>  	{ "raw"       , 0, NULL, 'r' },
>  	{ "pid"       , 1, NULL, 'p' },
>  	{ "file"      , 1, NULL, 'f' },
>  	{ "addr"      , 1, NULL, 'a' },
> +	{ "decode"    , 1, NULL, 'd' },
>  	{ "bits"      , 1, NULL, 'b' },
>  	{ "list"      , 0, NULL, 'l' },
>  	{ "list-each" , 0, NULL, 'L' },
> @@ -907,7 +926,7 @@ int main(int argc, char *argv[])
>  	page_size = getpagesize();
>  
>  	while ((c = getopt_long(argc, argv,
> -				"rp:f:a:b:lLNXxh", opts, NULL)) != -1) {
> +				"rp:f:a:b:d:lLNXxh", opts, NULL)) != -1) {
>  		switch (c) {
>  		case 'r':
>  			opt_raw = 1;
> @@ -924,6 +943,9 @@ int main(int argc, char *argv[])
>  		case 'b':
>  			parse_bits_mask(optarg);
>  			break;
> +		case 'd':
> +			decode_flags_and_exit(optarg);
> +			break;
>  		case 'l':
>  			opt_list = 1;
>  			break;

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] page-types: decode flags directly from command line

[Alex Chiang]

Hi Fengguang,

* Wu Fengguang <fengguang.wu@intel.com>:
> On Wed, Nov 04, 2009 at 06:54:41AM +0800, Alex Chiang wrote:
> > Why is this useful? For instance, if you're using memory hotplug
> > and see this in /var/log/messages:
> > 
> > 	kernel: removing from LRU failed 3836dd0/1/1e00000000000400
> > 
> > It would be nice to decode those page flags without staring at
> > the source.
> 
> In fact it's more than decode - encoding is also possible with the
> _same_ code! So maybe "-d" and help message will not be all that
> appropriate.

I'm sorry, I don't understand this use case, so I'm not sure what
you're asking me to do.

You're saying that a use case would be something like:

	./page-types --encode referenced,mmap
	0x0000000000000004

?

If that's what you're asking for, I guess I'm not sure why that's
so useful, but then again, I'm a vm n00b so there are probably
lots of things I don't understand. ;)

> > Example usage and output:
> > 
> > linux-2.6/Documentation/vm$ ./page-types -d 0x1e00000000000400
> >              flags	page-count       MB  symbolic-flags			long-symbolic-flags
> > 0x1e00000000000400	         1        0  __________B_______________________buddy
> >              total	         1        0
> 
> The output is a bit redundant - so does the code. Could you simplify
> them a bit?

Well, the code is redundant, but add_page() / show_summary() is a
simple sequence.

In contrast, I think I'd have to modify walk_addr_ranges() and
maybe walk_pfn() to do something special when we don't really
want to do any address space walking, and simply want to
decode/encode some user input.

Maybe I don't understand you fully? Could you give me a better
idea of what you're looking for?

As for the output, I'm just reusing show_summary(). Maybe we
don't need the flags, page-count, and MB columns, but again, the
patch would be more intrusisive because we'd have to teach
show_summary() about the special case.

Anyway, I'm happy to make changes closer to what you're looking
for, but I'd like some more guidance as to what you're expecting.

Thanks,
/ac

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [PATCH] page-types: decode flags directly from command line

[Wu Fengguang]

Hi Alex,

It's easier to illustrate the idea with example and code :)

# Documentation/vm/page-types -d 0x10,anon
0x0000000000001010      ____D_______a_____________________      dirty,anonymous


+static void describe_flags(const char *optarg)
+{
+	uint64_t flags = parse_flag_names(optarg, 0);
+
+	printf("0x%016llx\t%s\t%s\n",
+	       (unsigned long long)flags,
+	       page_flag_name(flags),
+	       page_flag_longname(flags));
+}
 
+		case 'd':
+			opt_no_summary = 1;
+			describe_flags(optarg);
+			break;

Thanks,
Fengguang

On Thu, Nov 05, 2009 at 04:40:08AM +0800, Alex Chiang wrote:
> Hi Fengguang,
> 
> * Wu Fengguang <fengguang.wu@intel.com>:
> > On Wed, Nov 04, 2009 at 06:54:41AM +0800, Alex Chiang wrote:
> > > Why is this useful? For instance, if you're using memory hotplug
> > > and see this in /var/log/messages:
> > > 
> > > 	kernel: removing from LRU failed 3836dd0/1/1e00000000000400
> > > 
> > > It would be nice to decode those page flags without staring at
> > > the source.
> > 
> > In fact it's more than decode - encoding is also possible with the
> > _same_ code! So maybe "-d" and help message will not be all that
> > appropriate.
> 
> I'm sorry, I don't understand this use case, so I'm not sure what
> you're asking me to do.
> 
> You're saying that a use case would be something like:
> 
> 	./page-types --encode referenced,mmap
> 	0x0000000000000004
> 
> ?
> 
> If that's what you're asking for, I guess I'm not sure why that's
> so useful, but then again, I'm a vm n00b so there are probably
> lots of things I don't understand. ;)
> 
> > > Example usage and output:
> > > 
> > > linux-2.6/Documentation/vm$ ./page-types -d 0x1e00000000000400
> > >              flags	page-count       MB  symbolic-flags			long-symbolic-flags
> > > 0x1e00000000000400	         1        0  __________B_______________________buddy
> > >              total	         1        0
> > 
> > The output is a bit redundant - so does the code. Could you simplify
> > them a bit?
> 
> Well, the code is redundant, but add_page() / show_summary() is a
> simple sequence.
> 
> In contrast, I think I'd have to modify walk_addr_ranges() and
> maybe walk_pfn() to do something special when we don't really
> want to do any address space walking, and simply want to
> decode/encode some user input.
> 
> Maybe I don't understand you fully? Could you give me a better
> idea of what you're looking for?
> 
> As for the output, I'm just reusing show_summary(). Maybe we
> don't need the flags, page-count, and MB columns, but again, the
> patch would be more intrusisive because we'd have to teach
> show_summary() about the special case.
> 
> Anyway, I'm happy to make changes closer to what you're looking
> for, but I'd like some more guidance as to what you're expecting.
> 
> Thanks,
> /ac

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>